Abstracts

Rationale: Dense-array EEG using 256 channels could potentially localize epileptogenic foci and map brain functions in epilepsy patients more precisely. However, it is impossible to mark the interictal epileptiform discharges (IEDs) by visual inspection from all 256 channels. Hence, neurologists have to translate the 256-channel data to the more traditional 10-20 montage to review the EEG offline, which defeats the purpose of using a high-density EEG system. Additionally, EMG artifact can be recorded very easily by the dense-array EEG, obscuring IEDs when using currently available IED detection algorithms. In this study, we introduce a new algorithm that can automatically detect IEDs in 256-channel data in the presence of EMG artifact. Methods: In a pilot study, dense-array EEGs were recorded from three epilepsy patients with eyes closed via the EGI Geodesic EEG 400 system at a sampling rate of 1000 Hz for 30 minutes. EEG data were exported and pre-processed with Matlab programs developed in-house. The EEG was down-sampled to 250 Hz and filtered between 1 Hz and 70 Hz. The global field power (GFP) and entropy across all 256 channels were computed at each time point. GFP is sensitive to polarity changes across channels, while entropy can be employed to eliminate the contaminating effects of EMG artifact. IEDs were identified by auto-thresholding both GFP and entropy. Source localization of the IEDs was determined using an sLORETA inverse solution in Brainstorm (a Matlab-based program) and mapped onto each patient's individual MRI. Intracranial EEG and/or intraoperative ECoG recordings were utilized to evaluate the accuracy of the IED source localizations. Results: IEDs were successfully detected from patients' 30-min EEG recordings, all of which have been inspected and verified by neurologists. Some IEDs in close proximity to the previous resection from one patient with recurrent temporal lobe epilepsy were not obvious when recorded with a traditional 10-20 montage. For patients with significant lesions shown on MRI, the IED locations were determined to be on top of the lesions. These IED source localization results are in agreement with findings from intracranial EEG and/or intraoperative ECoG recordings. Conclusions: Our automated detection algorithm introduced in this study is very precise at identifying IEDs from 256-channel EEG recordings. It is highly effective at capturing IEDs that might not be obvious using a traditional 10-20 montage. EMG artifacts could be easily identified and had no effect on IED detections. Funding: N/A